Certain investigations on crop yield prediction using machine learning algorithms

Abstract

A significant portion of the Indian economy is based on newlineagriculture. Agriculture depends enormously on soil quality and climate, so it newlineis imperative to advance in this area. A prevalent challenge faced by Indian newlinefarmers is a lack of awareness regarding suitable crops aligned with their soil newlinerequirements, potentially impacting productivity. This problem may be solved newlinethrough precision agriculture. Crop Yield Prediction (CYP) stands as a newlineformidable task within the agricultural sector. Significant research within this newlinedomain has concentrated on utilizing machine learning algorithms to improve newlinethe precision of predicting crop yields. Crop yield (CY) is a multifaceted newlinevariable impacted by various factors, such as genotype, environment, and their newlineinteractions. CYP represents a substantial concern in agriculture. CY depends newlinemainly on weather conditions, soil nutrients, and temperature. newlineWith this view in mind, this thesis proposes methods for crop newlineyield prediction using Artificial Neural Networks (ANN) which are usually newlineused to predict the behaviour of complex non-linear models. As a result, this newlineresearch attempts to determine the correlations between climatic variables, soil newlinenutrients and CY with the available data. In ANN, three methods, Levenberg newlineMarquardt (LM) , Bayesian regularisation (BR) and scaled conjugate gradient newline(SCG) are used to train the neural network model (NN) and then compared to newlinedetermine prediction accuracy. The training performance measures, such as newlinethe mean squared error (MSE) and the correlation coefficient (R), were newlinedetermined to assess the ANN models that had been built. The experimental newlinestudy proves that LM training algorithms are better, while BR and SCG have newlineminimal performance. newline